Harnessing Real-World Data to Advance Pharmacometrics in Understudied Populations: Computational Strategies, Study Design, and Machine Learning Approaches

The presentation will offer an academic perspective on harnessing the vast potential of Real-World Data (RWD) to advance pharmacometrics for heterogeneous and often understudied populations. We will discuss the unique opportunities presented by large datasets both for hypothesis generation leading to scientific discovery and for validating hypotheses derived from experimental or clinical contexts. This includes examples of computational strategies for secondary data analysis, emphasizing rigorous study designs (e.g. cohort selection) coupled with statistical adjustment approaches (e.g. propensity scores) to enable robust inference, often applied alongside supervised machine learning. We will also explore novel methods (unsupervised learning, generative models) to uncover heterogeneity and subgroup effects. Case examples, including work in Alzheimer's Disease, will illustrate how these approaches can reveal population heterogeneity, highlight understudied populations, define important subgroups, and ultimately generate evidence that complements treatment advances and informs advances in clinical care.